Ethics of AI and Machine Learning

Sleek flying cars, sentient robots, and time travel–for decades, artificial intelligence has tempted us with its possibilities. But will AI, if left unchecked, simply exacerbate humanity's existing social issues?

Ever since the release of ChatGPT, AI has become a new buzzword in the media. Companies like Google, Apple, and Microsoft are breathlessly racing each other to roll out the best chatbot. Meanwhile, actors and musicians are speaking out, with the SAG actors' guild going on strike to negotiate the use of artificial intelligence in the industry. As a developing field, AI raises many unanswered questions about its regulation and potential societal impacts. 

Defined as "the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings," artificial intelligence was founded as an academic discipline in 1956 at a Dartmouth conference. It is a field of research that teaches machines to be aware of their environment to make efficient decisions and perform complex tasks. With this new intelligence, AI has the potential to revolutionize a vast range of fields, promising increased efficiency and profit. 

While many are excited by the potential developments AI can help with, it has become apparent that AI is also recreating the same racial discrimination that has plagued society for decades. 

Algorithms, which are "a set of instructions to be followed in calculations," are a crucial part of AI. As these complex rules continue to develop, they are replicating structural racism and discrimination.

Safiya Noble, author of Algorithms of Oppression, explained in an episode of NPR's Code Switch, "Because AI makes it possible for machines to learn from experience, that means AI is susceptible to the same bias of the humans it's simulating." 

At MIT, students explore how technology reinforces discriminatory design at the Schwarzman College of Computing and across other departments. "Even with the most benevolent technology, no matter how well-intentioned we are ethically, we are still inevitably being discriminatory,” said MIT student Teresa Gao.

One example of this discriminatory design is in natural language processing, or NLP, a subset of AI that allows computers to understand and create human language. However, gender, ethnic, and racial bias in NLP is a major concern. When language models learn from training data that contains implicit biases, they can reinforce discrimination. 

This discrimination is also seen in medical settings, with clinical algorithms showing racial bias toward patients. In 2019, a study found that an algorithm deemed that Black patients had to be much sicker than white patients to be recommended for identical care.  

A landmark study by the National Institute of Standards and Technology in 2019 showed that government facial recognition systems often misidentified people of color more than white people. The effect of this was that suspects were wrongly arrested, often being charged for crimes they did not commit. 

The term "racial justice" means seeking justice by combating structural oppression rather than focusing on fixing individual cases of racism. It is about making sure that people of color do not suffer the most due to a broken system. 

Glen Harris, president of Race Forward, highlights that this "imbalanced system makes all of us pay."

As AI chatbots continue to roll out, various organizations and government leaders are convening to draft AI bills and regulations. But according to the ACLU, regulatory efforts remain "race-blind," which means that the tech sector will not prevent harm caused by racial inequity. 

One of the first global standards on AI ethics, written in 2021, Unesco's "Recommendation on the Ethics of Artificial Intelligence," highlighted human rights but not racial justice. The standard contained four main values: protecting human rights and dignity, living in peaceful societies, ensuring diversity and inclusiveness, and flourishing environments and ecosystems. 

It will be difficult to protect the lives of marginalized communities if the tech sector doesn't take concrete steps to acknowledge that this new technology is biased and not objective. 

"Most of us agree that we want computing to work for social good, but which good? Whose good? Whose needs and values and worldviews are prioritized and whose are overlooked?” said Catherine D’Ignazio, an assistant professor of urban science and planning and director of the Data + Feminism Lab at MIT.

For too long, the lives of people of color have gone overlooked, minimized, and even threatened by our institutions. As AI technology develops and is implemented across society, it is a chance to acknowledge the importance of racial justice. First, by admitting that it must be a priority, and then by taking proactive steps to address bias in the data and algorithms.

We could take it one step further—isn't it possible to create technology that would promote more equality in the world? 

Policymakers and CEOs will most likely need an incentive to find this work worthwhile. Historically, appealing to a person's sense of morality has not been enough to make them take racial justice seriously. It may be more effective to remind humanity how harming marginalized communities ultimately harms everyone by creating unsustainable systems. 

Ultimately, scholars, politicians, and engineers must work together to address the many racial justice issues posed by the development of artificial intelligence so that it can actually create the utopian vision so many creators use to promote it.